摘要
针对国内荧光手术导航设备中单一红外成像存在的组织细节缺失、清晰度不足及不适应人眼视觉习惯等问题,提出了一种基于轻量级APWnet卷积神经网络的实时红外与可见光图像融合方法。该网络融合了IFCNN与Mobilenet的设计理念,并引入SEnet通道注意力机制,采用双支路结构,深度可分离卷积结合跳跃连接来提取多尺度特征。同时减少卷积层通道数量优化网络结构,使其参数量减少至原APWnet的14.16%,前向传播计算量降低85.04%,从而满足手术导航所需的实时处理要求。实验结果显示,改进后的轻量级APWnet在网络运行速度上表现最佳,能够以31.22 FPS实现融合图像的输出,在多个图像质量评价指标中均表现出色,证实了该方法在保持图像细节、结构清晰度及改善视觉效果上的优势。
To address the problems such as the lack of tissue detail,insufficient clarity,and incompatibility with human visual habits in single infrared imaging used in domestic fluorescence surgical navigation devices,the paper presents a real-time infrared and visible light image fusion method based on a lightweight APWnet convolutional neural network.The network integrates the design concepts from IFCNN and Mobilenet,introduces the SENet channel attention mechanism,adopts a dual-branch structure,and combines depthwise separable convolutions with skip connections to extract multi-scale features.The network structure is optimized by reducing the number of channels in the convolutional layers,reducing its parameter count to 14.16%of the original APWnet and the forward pass computation by 85.04%,which meet the real-time processing requirements for surgical navigation.Experimental results show that the improved lightweight APWnet performs the best in terms of network operation speed,achieving fused image output at 31.22 FPS,and also performs excellently in multiple image quality evaluation metrics,which confirms its advantages in retaining image details,structural clarity,and enhancing visual quality.
作者
姚新雨
国蓉
朱鹏超
王晓东
何镇安
YAO Xinyu;GUO Rong;ZHU Pengchao;WANG Xiaodong;HE Zhenan(School of Opto-electronical Engineering,Xi’an Technological University,Xi’an 710021,China;Shaanxi Medical Device Quality Testing Institute,Xi’an 712046,China;School of Mechanical Engineering,Xi’an Jiaotong University,Xi’an 710049,China;Xi’an Institute of Optics and Precision Mechanics of CAS,Xi’an 710119,China)
出处
《西安工业大学学报》
2025年第4期608-619,共12页
Journal of Xi’an Technological University
基金
陕西省重点研发计划一般项目(2024SF-YBXM-436)
西安市科技计划项目(23ZDCYJSGG0007-2022)
电子信息专硕联合培养基地建设一般项目(XAGDYJ220217)。
关键词
荧光手术导航
吲哚菁绿
图像融合
实时
轻量型
自适应像素加权网络
fluorescence surgical navigation
Indocyanine Green(ICG)
image fusion
real-time
lightweight
Adaptive Pixel Weighting net(APWnet)
